Web page context may be used to assist spoken language understanding (SLU) applications. In conventional systems, recognizing speech and natural understanding processing needs some sort of grammars or context information. In speech recognition this is usually domain specific rule based grammars, such as Speech Recognition Grammar Systems (SRGS), or statistical language models, such as n-grams. In some situations, however, incoming speech may not be categorized, so wide-ranging statistical models must be used rather than tightly focused ones that may better recognize and respond to spoken queries.